Loading...
1/*
2 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#ifndef __XFS_INODE_H__
19#define __XFS_INODE_H__
20
21struct posix_acl;
22struct xfs_dinode;
23struct xfs_inode;
24
25/*
26 * Fork identifiers.
27 */
28#define XFS_DATA_FORK 0
29#define XFS_ATTR_FORK 1
30
31/*
32 * The following xfs_ext_irec_t struct introduces a second (top) level
33 * to the in-core extent allocation scheme. These structs are allocated
34 * in a contiguous block, creating an indirection array where each entry
35 * (irec) contains a pointer to a buffer of in-core extent records which
36 * it manages. Each extent buffer is 4k in size, since 4k is the system
37 * page size on Linux i386 and systems with larger page sizes don't seem
38 * to gain much, if anything, by using their native page size as the
39 * extent buffer size. Also, using 4k extent buffers everywhere provides
40 * a consistent interface for CXFS across different platforms.
41 *
42 * There is currently no limit on the number of irec's (extent lists)
43 * allowed, so heavily fragmented files may require an indirection array
44 * which spans multiple system pages of memory. The number of extents
45 * which would require this amount of contiguous memory is very large
46 * and should not cause problems in the foreseeable future. However,
47 * if the memory needed for the contiguous array ever becomes a problem,
48 * it is possible that a third level of indirection may be required.
49 */
50typedef struct xfs_ext_irec {
51 xfs_bmbt_rec_host_t *er_extbuf; /* block of extent records */
52 xfs_extnum_t er_extoff; /* extent offset in file */
53 xfs_extnum_t er_extcount; /* number of extents in page/block */
54} xfs_ext_irec_t;
55
56/*
57 * File incore extent information, present for each of data & attr forks.
58 */
59#define XFS_IEXT_BUFSZ 4096
60#define XFS_LINEAR_EXTS (XFS_IEXT_BUFSZ / (uint)sizeof(xfs_bmbt_rec_t))
61#define XFS_INLINE_EXTS 2
62#define XFS_INLINE_DATA 32
63typedef struct xfs_ifork {
64 int if_bytes; /* bytes in if_u1 */
65 int if_real_bytes; /* bytes allocated in if_u1 */
66 struct xfs_btree_block *if_broot; /* file's incore btree root */
67 short if_broot_bytes; /* bytes allocated for root */
68 unsigned char if_flags; /* per-fork flags */
69 unsigned char if_ext_max; /* max # of extent records */
70 union {
71 xfs_bmbt_rec_host_t *if_extents;/* linear map file exts */
72 xfs_ext_irec_t *if_ext_irec; /* irec map file exts */
73 char *if_data; /* inline file data */
74 } if_u1;
75 union {
76 xfs_bmbt_rec_host_t if_inline_ext[XFS_INLINE_EXTS];
77 /* very small file extents */
78 char if_inline_data[XFS_INLINE_DATA];
79 /* very small file data */
80 xfs_dev_t if_rdev; /* dev number if special */
81 uuid_t if_uuid; /* mount point value */
82 } if_u2;
83} xfs_ifork_t;
84
85/*
86 * Inode location information. Stored in the inode and passed to
87 * xfs_imap_to_bp() to get a buffer and dinode for a given inode.
88 */
89struct xfs_imap {
90 xfs_daddr_t im_blkno; /* starting BB of inode chunk */
91 ushort im_len; /* length in BBs of inode chunk */
92 ushort im_boffset; /* inode offset in block in bytes */
93};
94
95/*
96 * This is the xfs in-core inode structure.
97 * Most of the on-disk inode is embedded in the i_d field.
98 *
99 * The extent pointers/inline file space, however, are managed
100 * separately. The memory for this information is pointed to by
101 * the if_u1 unions depending on the type of the data.
102 * This is used to linearize the array of extents for fast in-core
103 * access. This is used until the file's number of extents
104 * surpasses XFS_MAX_INCORE_EXTENTS, at which point all extent pointers
105 * are accessed through the buffer cache.
106 *
107 * Other state kept in the in-core inode is used for identification,
108 * locking, transactional updating, etc of the inode.
109 *
110 * Generally, we do not want to hold the i_rlock while holding the
111 * i_ilock. Hierarchy is i_iolock followed by i_rlock.
112 *
113 * xfs_iptr_t contains all the inode fields up to and including the
114 * i_mnext and i_mprev fields, it is used as a marker in the inode
115 * chain off the mount structure by xfs_sync calls.
116 */
117
118typedef struct xfs_ictimestamp {
119 __int32_t t_sec; /* timestamp seconds */
120 __int32_t t_nsec; /* timestamp nanoseconds */
121} xfs_ictimestamp_t;
122
123/*
124 * NOTE: This structure must be kept identical to struct xfs_dinode
125 * in xfs_dinode.h except for the endianness annotations.
126 */
127typedef struct xfs_icdinode {
128 __uint16_t di_magic; /* inode magic # = XFS_DINODE_MAGIC */
129 __uint16_t di_mode; /* mode and type of file */
130 __int8_t di_version; /* inode version */
131 __int8_t di_format; /* format of di_c data */
132 __uint16_t di_onlink; /* old number of links to file */
133 __uint32_t di_uid; /* owner's user id */
134 __uint32_t di_gid; /* owner's group id */
135 __uint32_t di_nlink; /* number of links to file */
136 __uint16_t di_projid_lo; /* lower part of owner's project id */
137 __uint16_t di_projid_hi; /* higher part of owner's project id */
138 __uint8_t di_pad[6]; /* unused, zeroed space */
139 __uint16_t di_flushiter; /* incremented on flush */
140 xfs_ictimestamp_t di_atime; /* time last accessed */
141 xfs_ictimestamp_t di_mtime; /* time last modified */
142 xfs_ictimestamp_t di_ctime; /* time created/inode modified */
143 xfs_fsize_t di_size; /* number of bytes in file */
144 xfs_drfsbno_t di_nblocks; /* # of direct & btree blocks used */
145 xfs_extlen_t di_extsize; /* basic/minimum extent size for file */
146 xfs_extnum_t di_nextents; /* number of extents in data fork */
147 xfs_aextnum_t di_anextents; /* number of extents in attribute fork*/
148 __uint8_t di_forkoff; /* attr fork offs, <<3 for 64b align */
149 __int8_t di_aformat; /* format of attr fork's data */
150 __uint32_t di_dmevmask; /* DMIG event mask */
151 __uint16_t di_dmstate; /* DMIG state info */
152 __uint16_t di_flags; /* random flags, XFS_DIFLAG_... */
153 __uint32_t di_gen; /* generation number */
154} xfs_icdinode_t;
155
156/*
157 * Flags for xfs_ichgtime().
158 */
159#define XFS_ICHGTIME_MOD 0x1 /* data fork modification timestamp */
160#define XFS_ICHGTIME_CHG 0x2 /* inode field change timestamp */
161
162/*
163 * Per-fork incore inode flags.
164 */
165#define XFS_IFINLINE 0x01 /* Inline data is read in */
166#define XFS_IFEXTENTS 0x02 /* All extent pointers are read in */
167#define XFS_IFBROOT 0x04 /* i_broot points to the bmap b-tree root */
168#define XFS_IFEXTIREC 0x08 /* Indirection array of extent blocks */
169
170/*
171 * Fork handling.
172 */
173
174#define XFS_IFORK_Q(ip) ((ip)->i_d.di_forkoff != 0)
175#define XFS_IFORK_BOFF(ip) ((int)((ip)->i_d.di_forkoff << 3))
176
177#define XFS_IFORK_PTR(ip,w) \
178 ((w) == XFS_DATA_FORK ? \
179 &(ip)->i_df : \
180 (ip)->i_afp)
181#define XFS_IFORK_DSIZE(ip) \
182 (XFS_IFORK_Q(ip) ? \
183 XFS_IFORK_BOFF(ip) : \
184 XFS_LITINO((ip)->i_mount))
185#define XFS_IFORK_ASIZE(ip) \
186 (XFS_IFORK_Q(ip) ? \
187 XFS_LITINO((ip)->i_mount) - XFS_IFORK_BOFF(ip) : \
188 0)
189#define XFS_IFORK_SIZE(ip,w) \
190 ((w) == XFS_DATA_FORK ? \
191 XFS_IFORK_DSIZE(ip) : \
192 XFS_IFORK_ASIZE(ip))
193#define XFS_IFORK_FORMAT(ip,w) \
194 ((w) == XFS_DATA_FORK ? \
195 (ip)->i_d.di_format : \
196 (ip)->i_d.di_aformat)
197#define XFS_IFORK_FMT_SET(ip,w,n) \
198 ((w) == XFS_DATA_FORK ? \
199 ((ip)->i_d.di_format = (n)) : \
200 ((ip)->i_d.di_aformat = (n)))
201#define XFS_IFORK_NEXTENTS(ip,w) \
202 ((w) == XFS_DATA_FORK ? \
203 (ip)->i_d.di_nextents : \
204 (ip)->i_d.di_anextents)
205#define XFS_IFORK_NEXT_SET(ip,w,n) \
206 ((w) == XFS_DATA_FORK ? \
207 ((ip)->i_d.di_nextents = (n)) : \
208 ((ip)->i_d.di_anextents = (n)))
209
210
211
212#ifdef __KERNEL__
213
214struct bhv_desc;
215struct xfs_buf;
216struct xfs_bmap_free;
217struct xfs_bmbt_irec;
218struct xfs_inode_log_item;
219struct xfs_mount;
220struct xfs_trans;
221struct xfs_dquot;
222
223typedef struct dm_attrs_s {
224 __uint32_t da_dmevmask; /* DMIG event mask */
225 __uint16_t da_dmstate; /* DMIG state info */
226 __uint16_t da_pad; /* DMIG extra padding */
227} dm_attrs_t;
228
229typedef struct xfs_inode {
230 /* Inode linking and identification information. */
231 struct xfs_mount *i_mount; /* fs mount struct ptr */
232 struct xfs_dquot *i_udquot; /* user dquot */
233 struct xfs_dquot *i_gdquot; /* group dquot */
234
235 /* Inode location stuff */
236 xfs_ino_t i_ino; /* inode number (agno/agino)*/
237 struct xfs_imap i_imap; /* location for xfs_imap() */
238
239 /* Extent information. */
240 xfs_ifork_t *i_afp; /* attribute fork pointer */
241 xfs_ifork_t i_df; /* data fork */
242
243 /* Transaction and locking information. */
244 struct xfs_inode_log_item *i_itemp; /* logging information */
245 mrlock_t i_lock; /* inode lock */
246 mrlock_t i_iolock; /* inode IO lock */
247 struct completion i_flush; /* inode flush completion q */
248 atomic_t i_pincount; /* inode pin count */
249 wait_queue_head_t i_ipin_wait; /* inode pinning wait queue */
250 spinlock_t i_flags_lock; /* inode i_flags lock */
251 /* Miscellaneous state. */
252 unsigned short i_flags; /* see defined flags below */
253 unsigned char i_update_core; /* timestamps/size is dirty */
254 unsigned int i_delayed_blks; /* count of delay alloc blks */
255
256 xfs_icdinode_t i_d; /* most of ondisk inode */
257
258 xfs_fsize_t i_size; /* in-memory size */
259 xfs_fsize_t i_new_size; /* size when write completes */
260 atomic_t i_iocount; /* outstanding I/O count */
261
262 /* VFS inode */
263 struct inode i_vnode; /* embedded VFS inode */
264} xfs_inode_t;
265
266#define XFS_ISIZE(ip) S_ISREG((ip)->i_d.di_mode) ? \
267 (ip)->i_size : (ip)->i_d.di_size;
268
269/* Convert from vfs inode to xfs inode */
270static inline struct xfs_inode *XFS_I(struct inode *inode)
271{
272 return container_of(inode, struct xfs_inode, i_vnode);
273}
274
275/* convert from xfs inode to vfs inode */
276static inline struct inode *VFS_I(struct xfs_inode *ip)
277{
278 return &ip->i_vnode;
279}
280
281/*
282 * i_flags helper functions
283 */
284static inline void
285__xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
286{
287 ip->i_flags |= flags;
288}
289
290static inline void
291xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
292{
293 spin_lock(&ip->i_flags_lock);
294 __xfs_iflags_set(ip, flags);
295 spin_unlock(&ip->i_flags_lock);
296}
297
298static inline void
299xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
300{
301 spin_lock(&ip->i_flags_lock);
302 ip->i_flags &= ~flags;
303 spin_unlock(&ip->i_flags_lock);
304}
305
306static inline int
307__xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
308{
309 return (ip->i_flags & flags);
310}
311
312static inline int
313xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
314{
315 int ret;
316 spin_lock(&ip->i_flags_lock);
317 ret = __xfs_iflags_test(ip, flags);
318 spin_unlock(&ip->i_flags_lock);
319 return ret;
320}
321
322static inline int
323xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
324{
325 int ret;
326
327 spin_lock(&ip->i_flags_lock);
328 ret = ip->i_flags & flags;
329 if (ret)
330 ip->i_flags &= ~flags;
331 spin_unlock(&ip->i_flags_lock);
332 return ret;
333}
334
335/*
336 * Project quota id helpers (previously projid was 16bit only
337 * and using two 16bit values to hold new 32bit projid was chosen
338 * to retain compatibility with "old" filesystems).
339 */
340static inline prid_t
341xfs_get_projid(struct xfs_inode *ip)
342{
343 return (prid_t)ip->i_d.di_projid_hi << 16 | ip->i_d.di_projid_lo;
344}
345
346static inline void
347xfs_set_projid(struct xfs_inode *ip,
348 prid_t projid)
349{
350 ip->i_d.di_projid_hi = (__uint16_t) (projid >> 16);
351 ip->i_d.di_projid_lo = (__uint16_t) (projid & 0xffff);
352}
353
354/*
355 * Manage the i_flush queue embedded in the inode. This completion
356 * queue synchronizes processes attempting to flush the in-core
357 * inode back to disk.
358 */
359static inline void xfs_iflock(xfs_inode_t *ip)
360{
361 wait_for_completion(&ip->i_flush);
362}
363
364static inline int xfs_iflock_nowait(xfs_inode_t *ip)
365{
366 return try_wait_for_completion(&ip->i_flush);
367}
368
369static inline void xfs_ifunlock(xfs_inode_t *ip)
370{
371 complete(&ip->i_flush);
372}
373
374/*
375 * In-core inode flags.
376 */
377#define XFS_IRECLAIM 0x0001 /* started reclaiming this inode */
378#define XFS_ISTALE 0x0002 /* inode has been staled */
379#define XFS_IRECLAIMABLE 0x0004 /* inode can be reclaimed */
380#define XFS_INEW 0x0008 /* inode has just been allocated */
381#define XFS_IFILESTREAM 0x0010 /* inode is in a filestream directory */
382#define XFS_ITRUNCATED 0x0020 /* truncated down so flush-on-close */
383#define XFS_IDIRTY_RELEASE 0x0040 /* dirty release already seen */
384
385/*
386 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
387 * inode lookup. Thi prevents unintended behaviour on the new inode from
388 * ocurring.
389 */
390#define XFS_IRECLAIM_RESET_FLAGS \
391 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
392 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED | \
393 XFS_IFILESTREAM);
394
395/*
396 * Flags for inode locking.
397 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
398 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
399 */
400#define XFS_IOLOCK_EXCL (1<<0)
401#define XFS_IOLOCK_SHARED (1<<1)
402#define XFS_ILOCK_EXCL (1<<2)
403#define XFS_ILOCK_SHARED (1<<3)
404#define XFS_IUNLOCK_NONOTIFY (1<<4)
405
406#define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
407 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED)
408
409#define XFS_LOCK_FLAGS \
410 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
411 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
412 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
413 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
414 { XFS_IUNLOCK_NONOTIFY, "IUNLOCK_NONOTIFY" }
415
416
417/*
418 * Flags for lockdep annotations.
419 *
420 * XFS_LOCK_PARENT - for directory operations that require locking a
421 * parent directory inode and a child entry inode. The parent gets locked
422 * with this flag so it gets a lockdep subclass of 1 and the child entry
423 * lock will have a lockdep subclass of 0.
424 *
425 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
426 * inodes do not participate in the normal lock order, and thus have their
427 * own subclasses.
428 *
429 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
430 * with xfs_lock_inodes(). This flag is used as the starting subclass
431 * and each subsequent lock acquired will increment the subclass by one.
432 * So the first lock acquired will have a lockdep subclass of 4, the
433 * second lock will have a lockdep subclass of 5, and so on. It is
434 * the responsibility of the class builder to shift this to the correct
435 * portion of the lock_mode lockdep mask.
436 */
437#define XFS_LOCK_PARENT 1
438#define XFS_LOCK_RTBITMAP 2
439#define XFS_LOCK_RTSUM 3
440#define XFS_LOCK_INUMORDER 4
441
442#define XFS_IOLOCK_SHIFT 16
443#define XFS_IOLOCK_PARENT (XFS_LOCK_PARENT << XFS_IOLOCK_SHIFT)
444
445#define XFS_ILOCK_SHIFT 24
446#define XFS_ILOCK_PARENT (XFS_LOCK_PARENT << XFS_ILOCK_SHIFT)
447#define XFS_ILOCK_RTBITMAP (XFS_LOCK_RTBITMAP << XFS_ILOCK_SHIFT)
448#define XFS_ILOCK_RTSUM (XFS_LOCK_RTSUM << XFS_ILOCK_SHIFT)
449
450#define XFS_IOLOCK_DEP_MASK 0x00ff0000
451#define XFS_ILOCK_DEP_MASK 0xff000000
452#define XFS_LOCK_DEP_MASK (XFS_IOLOCK_DEP_MASK | XFS_ILOCK_DEP_MASK)
453
454#define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) >> XFS_IOLOCK_SHIFT)
455#define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) >> XFS_ILOCK_SHIFT)
456
457extern struct lock_class_key xfs_iolock_reclaimable;
458
459/*
460 * For multiple groups support: if S_ISGID bit is set in the parent
461 * directory, group of new file is set to that of the parent, and
462 * new subdirectory gets S_ISGID bit from parent.
463 */
464#define XFS_INHERIT_GID(pip) \
465 (((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
466 ((pip)->i_d.di_mode & S_ISGID))
467
468/*
469 * xfs_iget.c prototypes.
470 */
471int xfs_iget(struct xfs_mount *, struct xfs_trans *, xfs_ino_t,
472 uint, uint, xfs_inode_t **);
473void xfs_ilock(xfs_inode_t *, uint);
474int xfs_ilock_nowait(xfs_inode_t *, uint);
475void xfs_iunlock(xfs_inode_t *, uint);
476void xfs_ilock_demote(xfs_inode_t *, uint);
477int xfs_isilocked(xfs_inode_t *, uint);
478uint xfs_ilock_map_shared(xfs_inode_t *);
479void xfs_iunlock_map_shared(xfs_inode_t *, uint);
480void xfs_inode_free(struct xfs_inode *ip);
481
482/*
483 * xfs_inode.c prototypes.
484 */
485int xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t,
486 xfs_nlink_t, xfs_dev_t, prid_t, int,
487 struct xfs_buf **, boolean_t *, xfs_inode_t **);
488
489uint xfs_ip2xflags(struct xfs_inode *);
490uint xfs_dic2xflags(struct xfs_dinode *);
491int xfs_ifree(struct xfs_trans *, xfs_inode_t *,
492 struct xfs_bmap_free *);
493int xfs_itruncate_extents(struct xfs_trans **, struct xfs_inode *,
494 int, xfs_fsize_t);
495int xfs_itruncate_data(struct xfs_trans **, struct xfs_inode *,
496 xfs_fsize_t);
497int xfs_iunlink(struct xfs_trans *, xfs_inode_t *);
498
499void xfs_iext_realloc(xfs_inode_t *, int, int);
500void xfs_iunpin_wait(xfs_inode_t *);
501int xfs_iflush(xfs_inode_t *, uint);
502void xfs_lock_inodes(xfs_inode_t **, int, uint);
503void xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint);
504
505void xfs_synchronize_times(xfs_inode_t *);
506void xfs_mark_inode_dirty(xfs_inode_t *);
507void xfs_mark_inode_dirty_sync(xfs_inode_t *);
508
509#define IHOLD(ip) \
510do { \
511 ASSERT(atomic_read(&VFS_I(ip)->i_count) > 0) ; \
512 ihold(VFS_I(ip)); \
513 trace_xfs_ihold(ip, _THIS_IP_); \
514} while (0)
515
516#define IRELE(ip) \
517do { \
518 trace_xfs_irele(ip, _THIS_IP_); \
519 iput(VFS_I(ip)); \
520} while (0)
521
522#endif /* __KERNEL__ */
523
524/*
525 * Flags for xfs_iget()
526 */
527#define XFS_IGET_CREATE 0x1
528#define XFS_IGET_UNTRUSTED 0x2
529
530int xfs_inotobp(struct xfs_mount *, struct xfs_trans *,
531 xfs_ino_t, struct xfs_dinode **,
532 struct xfs_buf **, int *, uint);
533int xfs_itobp(struct xfs_mount *, struct xfs_trans *,
534 struct xfs_inode *, struct xfs_dinode **,
535 struct xfs_buf **, uint);
536int xfs_iread(struct xfs_mount *, struct xfs_trans *,
537 struct xfs_inode *, uint);
538void xfs_dinode_to_disk(struct xfs_dinode *,
539 struct xfs_icdinode *);
540void xfs_idestroy_fork(struct xfs_inode *, int);
541void xfs_idata_realloc(struct xfs_inode *, int, int);
542void xfs_iroot_realloc(struct xfs_inode *, int, int);
543int xfs_iread_extents(struct xfs_trans *, struct xfs_inode *, int);
544int xfs_iextents_copy(struct xfs_inode *, xfs_bmbt_rec_t *, int);
545
546xfs_bmbt_rec_host_t *xfs_iext_get_ext(xfs_ifork_t *, xfs_extnum_t);
547void xfs_iext_insert(xfs_inode_t *, xfs_extnum_t, xfs_extnum_t,
548 xfs_bmbt_irec_t *, int);
549void xfs_iext_add(xfs_ifork_t *, xfs_extnum_t, int);
550void xfs_iext_add_indirect_multi(xfs_ifork_t *, int, xfs_extnum_t, int);
551void xfs_iext_remove(xfs_inode_t *, xfs_extnum_t, int, int);
552void xfs_iext_remove_inline(xfs_ifork_t *, xfs_extnum_t, int);
553void xfs_iext_remove_direct(xfs_ifork_t *, xfs_extnum_t, int);
554void xfs_iext_remove_indirect(xfs_ifork_t *, xfs_extnum_t, int);
555void xfs_iext_realloc_direct(xfs_ifork_t *, int);
556void xfs_iext_direct_to_inline(xfs_ifork_t *, xfs_extnum_t);
557void xfs_iext_inline_to_direct(xfs_ifork_t *, int);
558void xfs_iext_destroy(xfs_ifork_t *);
559xfs_bmbt_rec_host_t *xfs_iext_bno_to_ext(xfs_ifork_t *, xfs_fileoff_t, int *);
560xfs_ext_irec_t *xfs_iext_bno_to_irec(xfs_ifork_t *, xfs_fileoff_t, int *);
561xfs_ext_irec_t *xfs_iext_idx_to_irec(xfs_ifork_t *, xfs_extnum_t *, int *, int);
562void xfs_iext_irec_init(xfs_ifork_t *);
563xfs_ext_irec_t *xfs_iext_irec_new(xfs_ifork_t *, int);
564void xfs_iext_irec_remove(xfs_ifork_t *, int);
565void xfs_iext_irec_compact(xfs_ifork_t *);
566void xfs_iext_irec_compact_pages(xfs_ifork_t *);
567void xfs_iext_irec_compact_full(xfs_ifork_t *);
568void xfs_iext_irec_update_extoffs(xfs_ifork_t *, int, int);
569
570#define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
571
572#if defined(DEBUG)
573void xfs_inobp_check(struct xfs_mount *, struct xfs_buf *);
574#else
575#define xfs_inobp_check(mp, bp)
576#endif /* DEBUG */
577
578extern struct kmem_zone *xfs_ifork_zone;
579extern struct kmem_zone *xfs_inode_zone;
580extern struct kmem_zone *xfs_ili_zone;
581
582#endif /* __XFS_INODE_H__ */
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
4 * All Rights Reserved.
5 */
6#ifndef __XFS_INODE_H__
7#define __XFS_INODE_H__
8
9#include "xfs_inode_buf.h"
10#include "xfs_inode_fork.h"
11
12/*
13 * Kernel only inode definitions
14 */
15struct xfs_dinode;
16struct xfs_inode;
17struct xfs_buf;
18struct xfs_bmbt_irec;
19struct xfs_inode_log_item;
20struct xfs_mount;
21struct xfs_trans;
22struct xfs_dquot;
23
24typedef struct xfs_inode {
25 /* Inode linking and identification information. */
26 struct xfs_mount *i_mount; /* fs mount struct ptr */
27 struct xfs_dquot *i_udquot; /* user dquot */
28 struct xfs_dquot *i_gdquot; /* group dquot */
29 struct xfs_dquot *i_pdquot; /* project dquot */
30
31 /* Inode location stuff */
32 xfs_ino_t i_ino; /* inode number (agno/agino)*/
33 struct xfs_imap i_imap; /* location for xfs_imap() */
34
35 /* Extent information. */
36 struct xfs_ifork *i_afp; /* attribute fork pointer */
37 struct xfs_ifork *i_cowfp; /* copy on write extents */
38 struct xfs_ifork i_df; /* data fork */
39
40 /* Transaction and locking information. */
41 struct xfs_inode_log_item *i_itemp; /* logging information */
42 mrlock_t i_lock; /* inode lock */
43 mrlock_t i_mmaplock; /* inode mmap IO lock */
44 atomic_t i_pincount; /* inode pin count */
45
46 /*
47 * Bitsets of inode metadata that have been checked and/or are sick.
48 * Callers must hold i_flags_lock before accessing this field.
49 */
50 uint16_t i_checked;
51 uint16_t i_sick;
52
53 spinlock_t i_flags_lock; /* inode i_flags lock */
54 /* Miscellaneous state. */
55 unsigned long i_flags; /* see defined flags below */
56 uint64_t i_delayed_blks; /* count of delay alloc blks */
57
58 struct xfs_icdinode i_d; /* most of ondisk inode */
59
60 /* VFS inode */
61 struct inode i_vnode; /* embedded VFS inode */
62
63 /* pending io completions */
64 spinlock_t i_ioend_lock;
65 struct work_struct i_ioend_work;
66 struct list_head i_ioend_list;
67} xfs_inode_t;
68
69/* Convert from vfs inode to xfs inode */
70static inline struct xfs_inode *XFS_I(struct inode *inode)
71{
72 return container_of(inode, struct xfs_inode, i_vnode);
73}
74
75/* convert from xfs inode to vfs inode */
76static inline struct inode *VFS_I(struct xfs_inode *ip)
77{
78 return &ip->i_vnode;
79}
80
81/*
82 * For regular files we only update the on-disk filesize when actually
83 * writing data back to disk. Until then only the copy in the VFS inode
84 * is uptodate.
85 */
86static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
87{
88 if (S_ISREG(VFS_I(ip)->i_mode))
89 return i_size_read(VFS_I(ip));
90 return ip->i_d.di_size;
91}
92
93/*
94 * If this I/O goes past the on-disk inode size update it unless it would
95 * be past the current in-core inode size.
96 */
97static inline xfs_fsize_t
98xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
99{
100 xfs_fsize_t i_size = i_size_read(VFS_I(ip));
101
102 if (new_size > i_size || new_size < 0)
103 new_size = i_size;
104 return new_size > ip->i_d.di_size ? new_size : 0;
105}
106
107/*
108 * i_flags helper functions
109 */
110static inline void
111__xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
112{
113 ip->i_flags |= flags;
114}
115
116static inline void
117xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
118{
119 spin_lock(&ip->i_flags_lock);
120 __xfs_iflags_set(ip, flags);
121 spin_unlock(&ip->i_flags_lock);
122}
123
124static inline void
125xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
126{
127 spin_lock(&ip->i_flags_lock);
128 ip->i_flags &= ~flags;
129 spin_unlock(&ip->i_flags_lock);
130}
131
132static inline int
133__xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
134{
135 return (ip->i_flags & flags);
136}
137
138static inline int
139xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
140{
141 int ret;
142 spin_lock(&ip->i_flags_lock);
143 ret = __xfs_iflags_test(ip, flags);
144 spin_unlock(&ip->i_flags_lock);
145 return ret;
146}
147
148static inline int
149xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
150{
151 int ret;
152
153 spin_lock(&ip->i_flags_lock);
154 ret = ip->i_flags & flags;
155 if (ret)
156 ip->i_flags &= ~flags;
157 spin_unlock(&ip->i_flags_lock);
158 return ret;
159}
160
161static inline int
162xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned short flags)
163{
164 int ret;
165
166 spin_lock(&ip->i_flags_lock);
167 ret = ip->i_flags & flags;
168 if (!ret)
169 ip->i_flags |= flags;
170 spin_unlock(&ip->i_flags_lock);
171 return ret;
172}
173
174static inline prid_t
175xfs_get_initial_prid(struct xfs_inode *dp)
176{
177 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
178 return dp->i_d.di_projid;
179
180 return XFS_PROJID_DEFAULT;
181}
182
183static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
184{
185 return ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK;
186}
187
188/*
189 * Check if an inode has any data in the COW fork. This might be often false
190 * even for inodes with the reflink flag when there is no pending COW operation.
191 */
192static inline bool xfs_inode_has_cow_data(struct xfs_inode *ip)
193{
194 return ip->i_cowfp && ip->i_cowfp->if_bytes;
195}
196
197/*
198 * Return the buftarg used for data allocations on a given inode.
199 */
200#define xfs_inode_buftarg(ip) \
201 (XFS_IS_REALTIME_INODE(ip) ? \
202 (ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
203
204/*
205 * In-core inode flags.
206 */
207#define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
208#define XFS_ISTALE (1 << 1) /* inode has been staled */
209#define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
210#define __XFS_INEW_BIT 3 /* inode has just been allocated */
211#define XFS_INEW (1 << __XFS_INEW_BIT)
212#define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
213#define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
214#define __XFS_IFLOCK_BIT 7 /* inode is being flushed right now */
215#define XFS_IFLOCK (1 << __XFS_IFLOCK_BIT)
216#define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
217#define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
218#define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
219/*
220 * If this unlinked inode is in the middle of recovery, don't let drop_inode
221 * truncate and free the inode. This can happen if we iget the inode during
222 * log recovery to replay a bmap operation on the inode.
223 */
224#define XFS_IRECOVERY (1 << 11)
225#define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
226
227/*
228 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
229 * inode lookup. This prevents unintended behaviour on the new inode from
230 * ocurring.
231 */
232#define XFS_IRECLAIM_RESET_FLAGS \
233 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
234 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED)
235
236/*
237 * Synchronize processes attempting to flush the in-core inode back to disk.
238 */
239
240static inline int xfs_isiflocked(struct xfs_inode *ip)
241{
242 return xfs_iflags_test(ip, XFS_IFLOCK);
243}
244
245extern void __xfs_iflock(struct xfs_inode *ip);
246
247static inline int xfs_iflock_nowait(struct xfs_inode *ip)
248{
249 return !xfs_iflags_test_and_set(ip, XFS_IFLOCK);
250}
251
252static inline void xfs_iflock(struct xfs_inode *ip)
253{
254 if (!xfs_iflock_nowait(ip))
255 __xfs_iflock(ip);
256}
257
258static inline void xfs_ifunlock(struct xfs_inode *ip)
259{
260 ASSERT(xfs_isiflocked(ip));
261 xfs_iflags_clear(ip, XFS_IFLOCK);
262 smp_mb();
263 wake_up_bit(&ip->i_flags, __XFS_IFLOCK_BIT);
264}
265
266/*
267 * Flags for inode locking.
268 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
269 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
270 */
271#define XFS_IOLOCK_EXCL (1<<0)
272#define XFS_IOLOCK_SHARED (1<<1)
273#define XFS_ILOCK_EXCL (1<<2)
274#define XFS_ILOCK_SHARED (1<<3)
275#define XFS_MMAPLOCK_EXCL (1<<4)
276#define XFS_MMAPLOCK_SHARED (1<<5)
277
278#define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
279 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
280 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
281
282#define XFS_LOCK_FLAGS \
283 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
284 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
285 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
286 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
287 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
288 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
289
290
291/*
292 * Flags for lockdep annotations.
293 *
294 * XFS_LOCK_PARENT - for directory operations that require locking a
295 * parent directory inode and a child entry inode. IOLOCK requires nesting,
296 * MMAPLOCK does not support this class, ILOCK requires a single subclass
297 * to differentiate parent from child.
298 *
299 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
300 * inodes do not participate in the normal lock order, and thus have their
301 * own subclasses.
302 *
303 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
304 * with xfs_lock_inodes(). This flag is used as the starting subclass
305 * and each subsequent lock acquired will increment the subclass by one.
306 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
307 * limited to the subclasses we can represent via nesting. We need at least
308 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
309 * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
310 * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
311 * 8 subclasses supported by lockdep.
312 *
313 * This also means we have to number the sub-classes in the lowest bits of
314 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
315 * mask and we can't use bit-masking to build the subclasses. What a mess.
316 *
317 * Bit layout:
318 *
319 * Bit Lock Region
320 * 16-19 XFS_IOLOCK_SHIFT dependencies
321 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
322 * 24-31 XFS_ILOCK_SHIFT dependencies
323 *
324 * IOLOCK values
325 *
326 * 0-3 subclass value
327 * 4-7 unused
328 *
329 * MMAPLOCK values
330 *
331 * 0-3 subclass value
332 * 4-7 unused
333 *
334 * ILOCK values
335 * 0-4 subclass values
336 * 5 PARENT subclass (not nestable)
337 * 6 RTBITMAP subclass (not nestable)
338 * 7 RTSUM subclass (not nestable)
339 *
340 */
341#define XFS_IOLOCK_SHIFT 16
342#define XFS_IOLOCK_MAX_SUBCLASS 3
343#define XFS_IOLOCK_DEP_MASK 0x000f0000
344
345#define XFS_MMAPLOCK_SHIFT 20
346#define XFS_MMAPLOCK_NUMORDER 0
347#define XFS_MMAPLOCK_MAX_SUBCLASS 3
348#define XFS_MMAPLOCK_DEP_MASK 0x00f00000
349
350#define XFS_ILOCK_SHIFT 24
351#define XFS_ILOCK_PARENT_VAL 5
352#define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
353#define XFS_ILOCK_RTBITMAP_VAL 6
354#define XFS_ILOCK_RTSUM_VAL 7
355#define XFS_ILOCK_DEP_MASK 0xff000000
356#define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
357#define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
358#define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
359
360#define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
361 XFS_MMAPLOCK_DEP_MASK | \
362 XFS_ILOCK_DEP_MASK)
363
364#define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
365 >> XFS_IOLOCK_SHIFT)
366#define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
367 >> XFS_MMAPLOCK_SHIFT)
368#define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
369 >> XFS_ILOCK_SHIFT)
370
371/*
372 * Layouts are broken in the BREAK_WRITE case to ensure that
373 * layout-holders do not collide with local writes. Additionally,
374 * layouts are broken in the BREAK_UNMAP case to make sure the
375 * layout-holder has a consistent view of the file's extent map. While
376 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
377 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
378 * go idle.
379 */
380enum layout_break_reason {
381 BREAK_WRITE,
382 BREAK_UNMAP,
383};
384
385/*
386 * For multiple groups support: if S_ISGID bit is set in the parent
387 * directory, group of new file is set to that of the parent, and
388 * new subdirectory gets S_ISGID bit from parent.
389 */
390#define XFS_INHERIT_GID(pip) \
391 (((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
392 (VFS_I(pip)->i_mode & S_ISGID))
393
394int xfs_release(struct xfs_inode *ip);
395void xfs_inactive(struct xfs_inode *ip);
396int xfs_lookup(struct xfs_inode *dp, struct xfs_name *name,
397 struct xfs_inode **ipp, struct xfs_name *ci_name);
398int xfs_create(struct xfs_inode *dp, struct xfs_name *name,
399 umode_t mode, dev_t rdev, struct xfs_inode **ipp);
400int xfs_create_tmpfile(struct xfs_inode *dp, umode_t mode,
401 struct xfs_inode **ipp);
402int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
403 struct xfs_inode *ip);
404int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
405 struct xfs_name *target_name);
406int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name,
407 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
408 struct xfs_name *target_name,
409 struct xfs_inode *target_ip, unsigned int flags);
410
411void xfs_ilock(xfs_inode_t *, uint);
412int xfs_ilock_nowait(xfs_inode_t *, uint);
413void xfs_iunlock(xfs_inode_t *, uint);
414void xfs_ilock_demote(xfs_inode_t *, uint);
415int xfs_isilocked(xfs_inode_t *, uint);
416uint xfs_ilock_data_map_shared(struct xfs_inode *);
417uint xfs_ilock_attr_map_shared(struct xfs_inode *);
418
419uint xfs_ip2xflags(struct xfs_inode *);
420int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
421int xfs_itruncate_extents_flags(struct xfs_trans **,
422 struct xfs_inode *, int, xfs_fsize_t, int);
423void xfs_iext_realloc(xfs_inode_t *, int, int);
424
425int xfs_log_force_inode(struct xfs_inode *ip);
426void xfs_iunpin_wait(xfs_inode_t *);
427#define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
428
429int xfs_iflush_cluster(struct xfs_buf *);
430void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
431 struct xfs_inode *ip1, uint ip1_mode);
432
433xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip);
434xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip);
435
436int xfs_dir_ialloc(struct xfs_trans **, struct xfs_inode *, umode_t,
437 xfs_nlink_t, dev_t, prid_t,
438 struct xfs_inode **);
439
440static inline int
441xfs_itruncate_extents(
442 struct xfs_trans **tpp,
443 struct xfs_inode *ip,
444 int whichfork,
445 xfs_fsize_t new_size)
446{
447 return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
448}
449
450/* from xfs_file.c */
451enum xfs_prealloc_flags {
452 XFS_PREALLOC_SET = (1 << 1),
453 XFS_PREALLOC_CLEAR = (1 << 2),
454 XFS_PREALLOC_SYNC = (1 << 3),
455 XFS_PREALLOC_INVISIBLE = (1 << 4),
456};
457
458int xfs_update_prealloc_flags(struct xfs_inode *ip,
459 enum xfs_prealloc_flags flags);
460int xfs_break_layouts(struct inode *inode, uint *iolock,
461 enum layout_break_reason reason);
462
463/* from xfs_iops.c */
464extern void xfs_setup_inode(struct xfs_inode *ip);
465extern void xfs_setup_iops(struct xfs_inode *ip);
466extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
467
468/*
469 * When setting up a newly allocated inode, we need to call
470 * xfs_finish_inode_setup() once the inode is fully instantiated at
471 * the VFS level to prevent the rest of the world seeing the inode
472 * before we've completed instantiation. Otherwise we can do it
473 * the moment the inode lookup is complete.
474 */
475static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
476{
477 xfs_iflags_clear(ip, XFS_INEW);
478 barrier();
479 unlock_new_inode(VFS_I(ip));
480 wake_up_bit(&ip->i_flags, __XFS_INEW_BIT);
481}
482
483static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
484{
485 xfs_setup_inode(ip);
486 xfs_setup_iops(ip);
487 xfs_finish_inode_setup(ip);
488}
489
490void xfs_irele(struct xfs_inode *ip);
491
492extern struct kmem_zone *xfs_inode_zone;
493
494/* The default CoW extent size hint. */
495#define XFS_DEFAULT_COWEXTSZ_HINT 32
496
497int xfs_iunlink_init(struct xfs_perag *pag);
498void xfs_iunlink_destroy(struct xfs_perag *pag);
499
500void xfs_end_io(struct work_struct *work);
501
502int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
503void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
504
505#endif /* __XFS_INODE_H__ */